Method and device for monitoring the operating process in a reactor

Abstract

The application discloses a kind of reaction kettle inside operation process monitoring method and device, server method includes: according to preset period, through 5G explosion-proof camera pre-deployed in reaction kettle, the video sequence to be analyzed in current period in kettle is collected;The video sequence to be analyzed in kettle is input into pre-trained chemical product visual analysis model, and the current product visual information corresponding to the video data to be analyzed in kettle is output;According to preset process operation node information, current product visual information, the target process operation node to be executed in current period is analyzed;In the case where target process operation node is not consistent with the historical process operation node of last historical period, operation process prompt information of target process operation node is created, and sent to client end.Therefore, by using the embodiment of the present application, the reaction kettle can be monitored in real time, and the user can be reminded in time at different process operation nodes, the accuracy and timeliness of the operation process monitoring of the reaction kettle are improved, and the stability of product quality is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computers, and in particular to a method and device for monitoring an operation process in a reaction kettle. BACKGROUND

[0002] In the chemical industry, a reaction kettle is a core device for chemical reactions and production processes. Its internal operation process is extremely complex and critical, including the addition of multiple materials, the precise control of reaction conditions (such as temperature, pressure, stirring speed, etc.). The accuracy and timeliness of these operation processes directly affect product quality.

[0003] In the prior art, an operator determines a specific operation process by observing a transparent window of a reaction kettle. This method completely depends on the work experience and attention of the operator, and is prone to monitoring errors due to fatigue or negligence of the operator. In addition, manual observation is difficult to accurately capture and record key details in the operation process. This monitoring method is low in timeliness and accuracy, resulting in unstable product quality. SUMMARY

[0004] Embodiments of the present application provide a method and device for monitoring an operation process in a reaction kettle. To have a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not a general review, nor does it determine the key / important components or delineate the protection scope of these embodiments. Its only purpose is to present some concepts in a simple form as a prelude to the detailed description that follows.

[0005] In a first aspect, the embodiments of the present application provide a method for monitoring an operation process in a reaction kettle, applied to a server, and the method comprises:

[0006] Collecting, by a 5G explosion-proof camera pre-deployed in the reaction kettle, a to-be-analyzed in-kettle video sequence in a current period according to a preset period;

[0007] Inputting the to-be-analyzed in-kettle video sequence into a pre-trained chemical product visual analysis model, and outputting current product visual information corresponding to the to-be-analyzed in-kettle video data;

[0008] According to preset process operation node information and the current product visual information, analyzing a target process operation node to be executed in the current period; the preset process operation node information is used to reflect operation steps, chemical product visual description parameters, and chemical reaction conditions to be executed in different process operation nodes;

[0009] In the case where the target process operation node is inconsistent with a historical process operation node of a previous historical period, creating operation process prompt information of the target process operation node, and sending the operation process prompt information to a client.

[0010] Secondly, embodiments of this application provide an operation process monitoring device inside a reactor, the device comprising:

[0011] The video acquisition module is used to acquire the video sequence inside the reactor to be analyzed within the current cycle using a 5G explosion-proof camera pre-deployed in the reactor according to a preset cycle;

[0012] The product visual information output module is used to input the video sequence inside the vessel to be analyzed into a pre-trained chemical product visual analysis model and output the current product visual information corresponding to the video data inside the vessel to be analyzed.

[0013] The process operation node analysis module is used to analyze the target process operation nodes that need to be executed in the current cycle based on preset process operation node information and current product visual information. The preset process operation node information is used to reflect the operation steps, chemical product visual description parameters and chemical reaction conditions that need to be executed in different process operation nodes.

[0014] The operation process prompt information creation module is used to create operation process prompt information for the target process operation node and send it to the client when the target process operation node is inconsistent with the historical process operation node of the previous historical cycle.

[0015] The technical solutions provided in this application embodiment may include the following beneficial effects:

[0016] In this embodiment, on the one hand, by using a pre-trained chemical product visual analysis model to output the current product visual information, the deep learning-based automated analysis method can accurately identify key visual features such as material state, color changes, and sediment area ratio within the reactor, avoiding the subjectivity and instability of manual observation, improving the accuracy and timeliness of reactor operation monitoring, and ensuring stable product quality. On the other hand, by analyzing the product visual information and preset process operation node information in real time within the current cycle, the system can quickly determine the target process operation node to be executed. If the target process operation node is inconsistent with the historical process operation node of the previous cycle, the system will immediately generate operation process prompts and send them to the client, reminding the operator to adjust the operation steps in time. The automated real-time analysis and reminder mechanism can ensure that the operator receives accurate operation prompts at the first time, avoiding production problems caused by delayed operation, thereby significantly improving the timeliness of monitoring.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] Figure 1 This is a schematic flowchart of a method for monitoring the operation process inside a reactor, provided in an embodiment of this application.

[0020] Figure 2 This is a schematic diagram of an image frame captured in a real-world scenario by a 5G explosion-proof camera, as provided in an embodiment of this application.

[0021] Figure 3 This is a schematic diagram of a UI interface for providing prompt information according to an embodiment of this application;

[0022] Figure 4 This is a schematic block diagram of a node consistency determination process provided in an embodiment of this application;

[0023] Figure 5 This is a flowchart illustrating a model training method for a visual analysis model of chemical products provided in an embodiment of this application.

[0024] Figure 6 This is a schematic diagram of the structure of an operation monitoring device inside a reactor provided in an embodiment of this application;

[0025] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0026] The following description and accompanying drawings fully illustrate specific embodiments of this application to enable those skilled in the art to practice them.

[0027] It should be understood that the described embodiments are merely some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0028] In the following description, when referring to the accompanying drawings, the same numbers in different drawings denote the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0029] In the description of this application, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances. Furthermore, in the description of this application, unless otherwise stated, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.

[0030] Currently, operators determine the specific operating procedures by observing the transparent viewing window of the reactor.

[0031] The applicant recognizes that this method relies entirely on the operator's experience and attention, making it susceptible to monitoring errors due to operator fatigue or negligence. Furthermore, manual observation struggles to accurately capture and record critical details during operation. The timeliness and accuracy of this monitoring method are low, leading to inconsistent product quality.

[0032] To address the aforementioned problems, this application provides a method and apparatus for monitoring the operation process within a reactor, thereby resolving the issues present in the related technical problems. In the embodiments of this application, on one hand, by employing a pre-trained chemical product visual analysis model to output the current product visual information, the automated analysis method based on deep learning can accurately identify key visual features within the reactor, such as material state, color changes, and the proportion of precipitate area. This avoids the subjectivity and instability of manual observation, improves the accuracy and timeliness of monitoring the reactor operation process, and ensures stable product quality. On the other hand, by analyzing the product visual information and preset process operation node information within the current cycle in real time, the system can quickly determine the target process operation node to be executed. If the target process operation node is inconsistent with the historical process operation node of the previous cycle, the system will immediately generate operation process prompts and send them to the client, reminding the operator to adjust the operation steps in a timely manner. The automated real-time analysis and reminder mechanism ensures that the operator receives accurate operation prompts at the first opportunity, avoiding production problems caused by delayed operations, thereby significantly improving the timeliness of monitoring. The following describes the process in detail using exemplary embodiments.

[0033] The following will be combined with the appendix Figure 1 -Appendix Figure 5 This application provides a detailed description of the method for monitoring the operation process within a reactor, as provided in the embodiments of this application. This method can be implemented using a computer program and can run on an operation process monitoring device within a reactor based on the von Neumann architecture. This computer program can be integrated into an application or run as a standalone utility application.

[0034] Please see Figure 1 This document provides a flowchart illustrating a method for monitoring the operation process within a reactor, applicable to a server-side application. For example... Figure 1 As shown, the method in this application embodiment includes the following steps:

[0035] S101, according to a preset cycle, collects the video sequence inside the reactor to be analyzed within the current cycle using a 5G explosion-proof camera pre-deployed in the reactor;

[0036] The preset period is a pre-defined time interval used to determine when to collect video sequences from inside the reactor. For example, it could be every 30 seconds, every minute, or every 5 minutes. The specific setting of the preset period can be customized based on the actual business scenario. A reactor is a commonly used container in chemical production for carrying out chemical reactions. A 5G explosion-proof camera is an image acquisition device capable of obtaining clear video streams in harsh production environments. Facing the complex environment inside a reactor, such as high temperature, high humidity, and high humidity, online real-time quality visualization and analysis of products presents certain challenges. By deploying a 5G explosion-proof camera inside the reactor, which features customized cooling back-blowing air, circulating cooling water, support for variable zoom modes, and explosion-proof functions, the image frames captured by this camera in the actual production environment are as follows: Figure 2 As shown. The video sequence inside the vessel to be analyzed in the current period is the video sequence inside the vessel within a preset time period collected from the current time when the end time of the previous monitoring and the current time are equal to the preset period.

[0037] In some embodiments of this application, 5G explosion-proof cameras are deployed at key locations inside the reactor, such as the reactor opening. After the cameras are activated, the server collects video streams through the pre-deployed 5G explosion-proof cameras in the reactor according to a preset period, which serves as the video sequence inside the reactor to be analyzed in the current period.

[0038] S102, input the video sequence inside the vessel to be analyzed into the pre-trained chemical product visual analysis model, and output the current product visual information corresponding to the video data inside the vessel to be analyzed;

[0039] The video sequence contains visual information such as the color of materials inside the reactor, the distribution of precipitates, the frequency of bubble formation, and the stirring status. The pre-trained chemical product visual analysis model is a trained deep learning model specifically designed to analyze video data from the reactor and extract visual information about the chemical products. This model automatically identifies and extracts key visual features from the video, such as material color, precipitate area ratio, and bubble formation frequency. The current product visual information is the visual feature information about the chemical products inside the reactor extracted from the video sequence of the reactor to be analyzed.

[0040] In some embodiments of this application, the server loads a pre-trained chemical product visual analysis model into memory, extracts video frames frame by frame from the video sequence to be analyzed, and inputs each extracted video frame into the chemical product visual analysis model. The model extracts and analyzes the visual information in the video frames using a pre-trained weight file, and outputs the visual feature information corresponding to each video frame, such as the RGB value of the material color, the area ratio of the precipitate, and the frequency of bubble formation. Finally, the visual feature information of each video frame is integrated to obtain the overall visual information of the reactor in the current cycle, and the integrated visual information is output as the current product visual information. The current product visual information is shown in Table 1.

[0041] Table 1

[0042]

[0043] In some embodiments of this application, the specific process of generating a pre-trained visual analysis model for chemical products includes: acquiring historical video sequences collected for the historical operation process of a reactor, the historical video sequences being used to reflect the visual information of the entire historical reaction cycle of the reactor; analyzing the historical time period of each historical process operation node from the entire historical reaction cycle; extracting the operation process video segment corresponding to the historical time period of each historical process operation node from the historical video sequence, as the operation process video segment of each historical process operation node; constructing model training samples based on the operation process video segments of each historical process operation node; creating a visual analysis model for chemical products using a neural network; inputting the model training samples into the visual analysis model for chemical products, outputting the model loss value, and generating a pre-trained visual analysis model for chemical products when the loss value reaches its minimum.

[0044] S103, based on the preset process operation node information and the current product visual information, analyze the target process operation node to be executed in the current cycle; the preset process operation node information is used to reflect the operation steps to be executed in different process operation nodes, the chemical product visual description parameters, and the chemical reaction conditions.

[0045] The preset process operation node information describes the specific operational steps, corresponding chemical product visual description parameters, and chemical reaction conditions required for each process stage (operation node) during reactor operation. Operation steps include, for example, "add raw material A," "stir for 3 minutes," and "heat to 80°C." Chemical product visual description parameters include, for example, "material color is yellow," "precipitate area ratio is less than 5%," and "bubble generation frequency is 3-5 bubbles / second." Chemical reaction conditions include, for example, "temperature range is 70-80°C." The target process operation node is the specific process operation node required for the current cycle, determined by analyzing the current visual information within the reactor and the preset process operation node information.

[0046] The preset process operation node information is shown in Table 2.

[0047] Table 2

[0048]

[0049] In some embodiments of this application, the specific process of analyzing the target process operation node to be executed in the current cycle based on preset process operation node information and current product visual information includes: obtaining real-time reaction condition parameters of the reactor in the current cycle; combining the real-time reaction condition parameters and current product visual information to obtain multiple real-time process quantification parameters in the current cycle; combining the chemical product visual description parameters and chemical reaction conditions of each process operation node to obtain multiple original process quantification parameters of each process operation node; generating a comprehensive matching degree for each process operation node based on multiple real-time process quantification parameters and multiple original process quantification parameters of each process operation node; and selecting the process operation node with the highest comprehensive matching degree as the target process operation node to be executed in the current cycle.

[0050] Real-time reaction condition parameters refer to the real-time physical and chemical conditions within the reactor during the current cycle, such as temperature. These parameters reflect the actual operating status within the reactor and, combined with visual information, are used to analyze the current process operation node.

[0051] In this embodiment, by selecting the process operation node with the highest matching degree as the target operation node of the current cycle, the current state of the reactor can be accurately identified, and the operation steps can be dynamically adjusted, thereby improving the automation and accuracy of production, reducing errors caused by manual intervention, and ensuring the improvement of product quality and production efficiency.

[0052] In some embodiments of this application, the specific process of generating the comprehensive matching degree of each process operation node based on multiple real-time process quantization parameters and multiple original process quantization parameters of each process operation node includes: performing feature normalization on multiple real-time process quantization parameters to obtain a first normalized feature; creating a master node and associating the first normalized feature with the master node to obtain a master node feature; performing feature normalization on multiple original process quantization parameters of each process operation node to obtain a second normalized feature of each process operation node; creating a child node for each process operation node; associating the second normalized feature of each process operation node with the child node of each process operation node to obtain multiple child node features; quantifying the spatial distance between the master node feature and each child node feature to obtain the comprehensive matching degree of each process operation node.

[0053] In other embodiments of this application, multiple real-time process quantification parameters include the RGB value of the current material color, the current temperature value inside the reactor, the current area ratio of the precipitate, and the current bubble generation frequency; multiple original process quantification parameters include a preset color range, a preset temperature range, a preset precipitate area range, and a preset bubble generation frequency range.

[0054] Specifically, the process of generating the comprehensive matching degree of each process operation node based on multiple real-time process quantification parameters and multiple original process quantification parameters of each process operation node includes: calculating the color matching degree between the RGB value of the current material color and the preset color range of each process operation node; determining whether the current temperature value in the reactor is within the preset temperature range of each process operation node, and generating a temperature matching result; calculating the overlap between the current area ratio of the precipitate and the preset precipitate area range of each process operation node, and obtaining the precipitate area matching degree; determining whether the current bubble generation frequency is within the preset bubble generation frequency range of each process operation node, and generating a bubble frequency matching result; and calculating the matching degree of the color matching degree, temperature matching result, precipitate area matching degree, and bubble frequency matching result to obtain the comprehensive matching degree of each process operation node.

[0055] In one possible implementation, during the current cycle, the following real-time reaction condition parameters are obtained from the sensors in the reactor: temperature: 35°C, stirring speed: 50 RPM. Current product visual information is obtained from the video analysis model, including: material color RGB values: (240, 180, 120), sediment area ratio: 6.5%, and bubble generation frequency: 3.8 bubbles / second. Then, the real-time reaction condition parameters and the current product visual information are combined to obtain the real-time process quantification parameters for the current cycle: material color RGB values: (240, 180, 120), sediment area ratio: 6.5%, bubble generation frequency: 3.8 bubbles / second, temperature: 35°C, and stirring speed: 50 RPM. Examples of combining the original process quantification parameters are shown in Table 2. Taking node 2 in Table 2 as an example, the RGB values ​​of the material color are: current value (240, 180, 120), range (150-200, 100-150, 50-100), center value (175, 125, 75). At this time:

[0056] Color matching ;

[0057] Precipitate area percentage: current value 6.5%, range 5-10%, center value 7.5%;

[0058] Sedimentation area matching degree ;

[0059] Bubble generation frequency: current value 3.8 bubbles / second, range 2-4 bubbles / second, center value 3 bubbles / second;

[0060] Bubble frequency matching degree ;

[0061] Temperature: Current value Range 30-40 Central value ;

[0062] Temperature matching ;

[0063] Overall matching degree of node 2 = =0.43333. Following the above calculation principles, the overall matching degree of other nodes can be calculated. Node 1: 0.62; Node 3: 0.35. Based on the calculation results, Node 1 has the highest overall matching degree (0.62), therefore, the target process operation node to be executed in the current cycle is Node 1.

[0064] S104. If the target process operation node is inconsistent with the historical process operation node of the previous historical cycle, create operation process prompt information for the target process operation node and send it to the client.

[0065] The historical process operation nodes from the previous historical period are those process operation nodes that have been executed or determined within the previous monitoring period. Operation process prompts are system-generated prompts that notify operators or automated control systems of the necessary operational steps when the target process operation node changes. The client is the terminal device or system that receives these operation process prompts; it can be an operator's workstation, mobile device, or automated control system.

[0066] In one possible implementation, if the target process operation node is inconsistent with the historical process operation node of the previous historical cycle, an operation process prompt message for creating the target process operation node is sent to the client.

[0067] For example, the system compares the target process operation node of the current cycle (e.g., node 2) with the historical process operation node of the previous cycle (e.g., node 1). Since node 2 is inconsistent with node 1, it indicates that the operation steps to be performed have changed. At this point, operation process prompts are generated based on the target process operation node (node ​​2) and sent to the client. The client's prompts include, for example... Figure 3 As shown.

[0068] In another possible implementation, if the target process operation node is consistent with the historical process operation node of the previous historical cycle, the process continues to execute the step of collecting the video sequence of the reactor interior to be analyzed within the current cycle using a 5G explosion-proof camera pre-deployed in the reactor, according to a preset cycle. Since the target process operation node is consistent with the historical process operation node, the system continues to collect the video sequence of the reactor interior to be analyzed within the current cycle using the 5G explosion-proof camera at a preset cycle (every 30 seconds). This process is automated and requires no additional intervention.

[0069] In this embodiment, when the target process operation node is consistent with the historical process operation node, video sequence acquisition continues automatically without additional intervention. This not only improves the automation level of monitoring but also reduces unnecessary operation prompts, avoids operator fatigue and interference, and ensures the efficient operation of the monitoring system.

[0070] For example Figure 4 As shown, the process first determines whether the target process operation node is consistent with the historical process operation node of the previous historical cycle. If so, it continues to execute the step of collecting the video sequence inside the reactor to be analyzed in the current cycle using a 5G explosion-proof camera pre-deployed in the reactor according to a preset cycle. If not, it creates an operation process prompt message for the target process operation node and sends it to the client.

[0071] It should be noted that this application applies to complex chemical reaction processes, especially those involving the addition of multiple materials and precise control of reaction conditions (such as temperature, pressure, stirring speed, pH, etc.), but where the visual changes in the product's state are minimal at key stages of the product preparation process. Specifically, it is applicable to reactions with extremely high requirements for reaction condition control, such as most reactions related to the synthesis of polymer compounds in fine chemicals. Precise monitoring of the reaction process can improve the quality of the final product in terms of hardness, ductility, stability, and other aspects.

[0072] The system can analyze product visual information and preset process operation node information in real time within the current cycle to quickly determine the target process operation node to be executed. If the target process operation node is inconsistent with the historical process operation node of the previous cycle, the system will immediately generate operation process prompts and send them to the client to remind the operator to adjust the operation steps in time.

[0073] In this embodiment, on the one hand, by using a pre-trained chemical product visual analysis model to output the current product visual information, the deep learning-based automated analysis method can accurately identify key visual features such as material state, color changes, and sediment area ratio within the reactor, avoiding the subjectivity and instability of manual observation, improving the accuracy and timeliness of reactor operation monitoring, and ensuring stable product quality. On the other hand, by analyzing the product visual information and preset process operation node information in real time within the current cycle, the system can quickly determine the target process operation node to be executed. If the target process operation node is inconsistent with the historical process operation node of the previous cycle, the system will immediately generate operation process prompts and send them to the client, reminding the operator to adjust the operation steps in time. The automated real-time analysis and reminder mechanism can ensure that the operator receives accurate operation prompts at the first time, avoiding production problems caused by delayed operation, thereby significantly improving the timeliness of monitoring.

[0074] Please see Figure 5 This is a flowchart illustrating a model training method for a visual analysis model of chemical products, provided in this application embodiment. Figure 5 As shown, the method in this application embodiment may include the following steps:

[0075] S201, acquire historical video sequences collected from the historical operation process of the reactor. The historical video sequences are used to reflect the visual information of the entire historical reaction cycle of the reactor.

[0076] The historical video sequence records video data of the reactor throughout its historical operation, from start to finish. This includes visual information such as color changes of materials within the reactor, the formation and distribution of precipitates, the generation and dissipation of bubbles, and the stirring status. The entire historical reaction cycle refers to the entire timeframe from the start of feeding materials to the end of a single chemical reaction.

[0077] In this embodiment, during model training, a 5G explosion-proof camera pre-deployed inside the reactor is used to collect video data at a preset frame rate (e.g., 1 frame per second). This ensures the camera can cover key areas within the reactor, such as the stirrer, material inlet, and reaction zone. The collected video data is stored in a data storage system, such as cloud storage or a local server. Each video sequence should contain the complete reaction cycle, from the start of feeding to the end of cooling.

[0078] S202, analyze the historical time period of each historical process operation node from the entire historical reaction cycle;

[0079] In some embodiments of this application, the specific process of analyzing the historical time period of each historical process operation node from the entire historical reaction cycle includes: extracting frames from the historical video sequence in chronological order to obtain multiple original video frames, each carrying a timestamp; obtaining the historical operation steps and historical chemical product visual description parameters of each historical process operation node based on preset process operation node information; traversing and searching for video frames that match the historical operation steps and historical chemical product visual description parameters of each historical process operation node from the multiple original video frames to determine the node start frame and node end frame of each historical process operation node; and determining the historical time period between the timestamps carried by the node start frame and node end frame of each historical process operation node as the historical time period of each historical process operation node.

[0080] Specifically, the process of traversing multiple original video frames to find video frames that match the historical operation steps and historical chemical product visual description parameters of each historical process operation node includes: calculating the pixel change rate between adjacent video frames in multiple original video frames based on the inter-frame difference algorithm; traversing and obtaining original video frames with pixel change rates greater than a preset threshold to obtain a sequence of video frames to be analyzed; defining operation tool information and operation gesture information for each video frame to be analyzed in the sequence to obtain the first operation step information of each video frame to be analyzed; analyzing the first chemical product visual description information of each video frame to be analyzed in the sequence to be analyzed; performing correlation analysis between the first operation step information and the first chemical product visual description information of each video frame to be analyzed and the historical operation steps and historical chemical product visual description parameters of each historical process operation node to obtain the correlation analysis results between each video frame to be analyzed and each historical process operation node; based on the correlation analysis results, identifying the video frames to be analyzed that are related to each historical process operation node, thus obtaining the video frames that match the historical operation steps and historical chemical product visual description parameters of each historical process operation node.

[0081] S203, extract the operation process video segment corresponding to the historical time period of each historical process operation node from the historical video sequence, and use it as the operation process video segment of each historical process operation node.

[0082] In some embodiments of this application, after obtaining the historical time period of each historical process operation node, the timestamp carried by each video frame in the historical video sequence can be compared with the start and end timestamps of the historical time period of each historical process operation node to find the operation process video segments between the same timestamps, which are then used as the operation process video segments of each historical process operation node.

[0083] S204, Construct model training samples based on video clips of the operation process of each historical process operation node;

[0084] In some embodiments of this application, the specific process of constructing model training samples based on the operation process video clips of each historical process operation node includes: obtaining each video frame from the operation process video clips of each historical process operation node; calculating the historical RGB value of the material color, the historical area ratio of the sediment, and the historical bubble generation frequency in each video frame; using the historical RGB value, the historical area ratio of the sediment, and the historical bubble generation frequency as label data; and using the label data to annotate each video frame to obtain model training samples.

[0085] S205, uses neural networks to create visual analysis models for chemical products;

[0086] The visual analysis model for chemical products can be constructed using convolutional neural networks or other neural networks; no specific limitations are specified here.

[0087] In some embodiments of this application, the specific process of creating a visual analysis model for chemical products using a neural network includes: a convolutional neural network (CNN) can be used to construct the visual analysis model for chemical products. This CNN includes an input layer, a convolutional layer, a pooling layer, and a fully connected layer. The input layer receives preprocessed video frames. The convolutional layer extracts local features from the image through convolution operations. The pooling layer reduces the dimensionality of the features while retaining important information through pooling operations. The fully connected layer integrates the extracted features and outputs the final prediction result.

[0088] S206: Input the model training samples into the chemical product visual analysis model, output the model loss value, and generate the pre-trained chemical product visual analysis model when the loss value reaches the minimum.

[0089] In some embodiments of this application, model training samples are input into a chemical product visual analysis model, the model loss value is output, and a pre-trained chemical product visual analysis model is generated when the loss value reaches its minimum. Alternatively, if the model loss value has not reached its minimum, the model parameters are adjusted, and the step of inputting model training samples into the chemical product visual analysis model continues.

[0090] In this embodiment, on the one hand, by using a pre-trained chemical product visual analysis model to output the current product visual information, the deep learning-based automated analysis method can accurately identify key visual features such as material state, color changes, and sediment area ratio within the reactor, avoiding the subjectivity and instability of manual observation, improving the accuracy and timeliness of reactor operation monitoring, and ensuring stable product quality. On the other hand, by analyzing the product visual information and preset process operation node information in real time within the current cycle, the system can quickly determine the target process operation node to be executed. If the target process operation node is inconsistent with the historical process operation node of the previous cycle, the system will immediately generate operation process prompts and send them to the client, reminding the operator to adjust the operation steps in time. The automated real-time analysis and reminder mechanism can ensure that the operator receives accurate operation prompts at the first time, avoiding production problems caused by delayed operation, thereby significantly improving the timeliness of monitoring.

[0091] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0092] Please see Figure 6 This illustration shows a schematic diagram of an operation monitoring device inside a reactor provided in an exemplary embodiment of this application. This operation monitoring device inside the reactor can be implemented as all or part of an electronic device through software, hardware, or a combination of both. The device 1 includes a video acquisition module 10, a product visual information output module 20, a process operation node analysis module 30, and an operation process prompt information creation module 40.

[0093] The video acquisition module 10 is used to acquire the video sequence inside the reactor to be analyzed within the current cycle using a 5G explosion-proof camera pre-deployed in the reactor according to a preset cycle;

[0094] The product visual information output module 20 is used to input the video sequence inside the vessel to be analyzed into a pre-trained chemical product visual analysis model and output the current product visual information corresponding to the video data inside the vessel to be analyzed.

[0095] The process operation node analysis module 30 is used to analyze the target process operation node to be executed in the current cycle based on the preset process operation node information and the current product visual information; the preset process operation node information is used to reflect the operation steps to be executed in different process operation nodes, the chemical product visual description parameters and the chemical reaction conditions.

[0096] The operation process prompt information creation module 40 is used to create operation process prompt information for the target process operation node and send it to the client when the target process operation node is inconsistent with the historical process operation node of the previous historical cycle.

[0097] It should be noted that the above embodiments of the reactor operation process monitoring device are only illustrative examples of the above functional module division when implementing the reactor operation process monitoring method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the equipment can be divided into different functional modules to complete all or part of the functions described above. In addition, the reactor operation process monitoring device and the reactor operation process monitoring method embodiments provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments, which will not be repeated here.

[0098] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0099] In this embodiment, on the one hand, by using a pre-trained chemical product visual analysis model to output the current product visual information, the deep learning-based automated analysis method can accurately identify key visual features such as material state, color changes, and sediment area ratio within the reactor, avoiding the subjectivity and instability of manual observation, improving the accuracy and timeliness of reactor operation monitoring, and ensuring stable product quality. On the other hand, by analyzing the product visual information and preset process operation node information in real time within the current cycle, the system can quickly determine the target process operation node to be executed. If the target process operation node is inconsistent with the historical process operation node of the previous cycle, the system will immediately generate operation process prompts and send them to the client, reminding the operator to adjust the operation steps in time. The automated real-time analysis and reminder mechanism can ensure that the operator receives accurate operation prompts at the first time, avoiding production problems caused by delayed operation, thereby significantly improving the timeliness of monitoring.

[0100] This application also provides a computer-readable medium having program instructions stored thereon, which, when executed by a processor, implement the operation process monitoring method within the reactor provided in the above-described method embodiments.

[0101] This application also provides a computer program product containing instructions that, when run on a computer, cause the computer to execute the operation process monitoring method within the reactor of the above-described method embodiments.

[0102] Please see Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 7As shown, the electronic device 1000 may include: at least one processor 1001, at least one network interface 1004, a user interface 1003, a memory 1005, and at least one communication bus 1002.

[0103] The communication bus 1002 is used to realize the connection and communication between these components.

[0104] The user interface 1003 may include a display screen and a camera. Optionally, the user interface 1003 may also include a standard wired interface and a wireless interface.

[0105] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0106] The processor 1001 may include one or more processing cores. The processor 1001 connects to various parts within the electronic device 1000 using various interfaces and lines. It executes various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 1005, and by calling data stored in the memory 1005. Optionally, the processor 1001 may be implemented using at least one hardware form selected from Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 1001 may integrate one or more of the following: a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), and a modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content to be displayed on the screen; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip, without being integrated into the processor 1001.

[0107] The memory 1005 may include random access memory (RAM) or read-only memory. Optionally, the memory 1005 may include a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 1005 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 1005 may also be at least one storage system located remotely from the aforementioned processor 1001. Figure 7 As shown, the memory 1005, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for monitoring the operation process within the reactor.

[0108] exist Figure 7 In the illustrated electronic device 1000, the user interface 1003 is mainly used to provide an input interface for the user and to acquire user input data; while the processor 1001 can be used to call the operation process monitoring application stored in the memory 1005 within the reactor and specifically perform the following operations:

[0109] According to a preset cycle, the video sequence inside the reactor to be analyzed is collected by a 5G explosion-proof camera pre-deployed in the reactor during the current cycle;

[0110] The video sequence inside the vessel to be analyzed is input into a pre-trained chemical product visual analysis model, which outputs the current product visual information corresponding to the video data inside the vessel to be analyzed.

[0111] Based on the preset process operation node information and the current product visual information, analyze the target process operation node that needs to be executed in the current cycle; the preset process operation node information is used to reflect the operation steps, chemical product visual description parameters and chemical reaction conditions that need to be executed in different process operation nodes.

[0112] If the target process operation node is inconsistent with the historical process operation node of the previous historical cycle, the operation process prompt information for creating the target process operation node is sent to the client.

[0113] In one embodiment, when the processor 1001 analyzes the target process operation node to be executed in the current cycle based on preset process operation node information and current product visual information, it specifically performs the following operations:

[0114] Obtain the real-time reaction condition parameters of the reactor during the current cycle;

[0115] By combining real-time response condition parameters and current product visual information, multiple real-time process quantification parameters within the current cycle can be obtained.

[0116] By combining the visual description parameters of the chemical products and the chemical reaction conditions of each process operation node, multiple original process quantification parameters for each process operation node are obtained.

[0117] Based on multiple real-time process quantization parameters and multiple original process quantization parameters for each process operation node, a comprehensive matching degree for each process operation node is generated.

[0118] The process operation node with the highest overall matching degree will be selected as the target process operation node to be executed in the current cycle.

[0119] In one embodiment, when the processor 1001 generates the comprehensive matching degree for each process operation node based on multiple real-time process quantization parameters and multiple raw process quantization parameters for each process operation node, it specifically performs the following operations:

[0120] Multiple real-time process quantization parameters are normalized to obtain the first normalized feature;

[0121] Create a master node and associate the first normalized feature with the master node to obtain the master node feature;

[0122] For each process operation node, multiple original process quantization parameters are normalized to obtain the second normalized feature of each process operation node.

[0123] Create a child node for each process operation node;

[0124] The second normalized feature of each process operation node is associated with the child nodes of each process operation node to obtain multiple child node features;

[0125] The spatial distance between the features of the master node and the features of each child node is quantified to obtain the comprehensive matching degree of each process operation node.

[0126] In one embodiment, when the processor 1001 generates the comprehensive matching degree for each process operation node based on multiple real-time process quantization parameters and multiple raw process quantization parameters for each process operation node, it specifically performs the following operations:

[0127] Calculate the color matching degree between the RGB value of the current material color and the preset color range of each process operation node;

[0128] Determine whether the current temperature value inside the reactor is within the preset temperature range for each process operation node, and generate a temperature matching result;

[0129] Calculate the degree of overlap between the current area ratio of the precipitate and the preset precipitate area range for each process operation node to obtain the precipitate area matching degree;

[0130] Determine whether the current bubble generation frequency is within the preset bubble generation frequency range for each process operation node, and generate bubble frequency matching results;

[0131] Determine the color matching degree, temperature matching degree, sedimentation area matching degree, and bubble frequency matching degree.

[0132] The matching degree of color matching, temperature matching, sedimentation area matching, and bubble frequency matching is calculated to obtain the comprehensive matching degree of each process operation node.

[0133] In one embodiment, the processor 1001 also performs the following operations:

[0134] If the target process operation node is consistent with the historical process operation node of the previous historical cycle, continue to execute the step of collecting the video sequence inside the reactor to be analyzed in the current cycle by using a 5G explosion-proof camera pre-deployed in the reactor according to the preset cycle.

[0135] In one embodiment, when the processor 1001 executes the generation of a pre-trained visual analysis model for chemical products, it specifically performs the following operations:

[0136] Acquire historical video sequences collected from the historical operation process of the reactor. These historical video sequences are used to reflect the visual information of the entire historical reaction cycle of the reactor.

[0137] Analyze the historical time periods of each historical process operation node throughout the entire historical reaction cycle;

[0138] From the historical video sequence, extract the operation process video segment corresponding to the historical time period of each historical process operation node, and use it as the operation process video segment of each historical process operation node;

[0139] Model training samples are constructed based on video clips of the operation process at each historical process node.

[0140] A visual analysis model for chemical products is created using neural networks;

[0141] The training samples are input into the chemical product visual analysis model, the model loss value is output, and the pre-trained chemical product visual analysis model is generated when the loss value reaches the minimum.

[0142] In one embodiment, when the processor 1001 analyzes the historical time period of each historical process operation node throughout the entire historical reaction cycle, it specifically performs the following operations:

[0143] The historical video sequence is frame-by-frame extracted in chronological order to obtain multiple raw video frames, each carrying a timestamp.

[0144] Based on the preset process operation node information, obtain the historical operation steps and historical chemical product visual description parameters for each historical process operation node;

[0145] From multiple original video frames, we traverse and search for video frames that match the historical operation steps and historical chemical product visual description parameters of each historical process operation node, so as to determine the node start frame and node end frame of each historical process operation node.

[0146] The historical time period between the timestamps carried by the node start frame and the node end frame of each historical process operation node is determined as the historical time period of each historical process operation node.

[0147] In one embodiment, when the processor 1001 performs the following operations when iterates through multiple raw video frames to find video frames that match the historical operating steps and historical chemical product visual description parameters of each historical process operation node:

[0148] Based on the inter-frame difference algorithm, the pixel change rate between adjacent video frames in multiple original video frames is calculated;

[0149] The original video frames with pixel change rates greater than a preset threshold are traversed and obtained to obtain the video frame sequence to be analyzed.

[0150] For each video frame in the video frame sequence to be analyzed, define operation tool information and operation gesture information to obtain the first operation step information of each video frame to be analyzed.

[0151] Analyze the visual description information of the first chemical product in each video frame of the video frame sequence to be analyzed;

[0152] The correlation analysis of the first operation step information and the first chemical product visual description information of each video frame to be analyzed with the historical operation steps and historical chemical product visual description parameters of each historical process operation node is performed to obtain the correlation analysis results between each video frame to be analyzed and each historical process operation node.

[0153] Based on the correlation analysis results, the video frames to be analyzed that are related to each historical process operation node were identified, and the video frames that match the historical operation steps and historical chemical product visual description parameters of each historical process operation node were obtained.

[0154] In one embodiment, when the processor 1001 constructs model training samples based on video clips of the operation process of each historical process operation node, it specifically performs the following operations:

[0155] Each video frame is obtained from the video clips of the operation process at each historical process operation node;

[0156] Calculate the historical RGB values ​​of material color, the historical area ratio of sediment, and the historical bubble generation frequency in each video frame;

[0157] Historical RGB values, historical area ratio of sediment, and historical bubble generation frequency are used as label data;

[0158] Label data is used to annotate each video frame to obtain training samples for the model.

[0159] In this embodiment, on the one hand, by using a pre-trained chemical product visual analysis model to output the current product visual information, the deep learning-based automated analysis method can accurately identify key visual features such as material state, color changes, and sediment area ratio within the reactor, avoiding the subjectivity and instability of manual observation, improving the accuracy and timeliness of reactor operation monitoring, and ensuring stable product quality. On the other hand, by analyzing the product visual information and preset process operation node information in real time within the current cycle, the system can quickly determine the target process operation node to be executed. If the target process operation node is inconsistent with the historical process operation node of the previous cycle, the system will immediately generate operation process prompts and send them to the client, reminding the operator to adjust the operation steps in time. The automated real-time analysis and reminder mechanism can ensure that the operator receives accurate operation prompts at the first time, avoiding production problems caused by delayed operation, thereby significantly improving the timeliness of monitoring.

[0160] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program for monitoring the operation process inside the reactor can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The storage medium for the program for monitoring the operation process inside the reactor can be a magnetic disk, optical disk, read-only memory, or random access memory, etc.

[0161] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A method for monitoring the operation process inside a reactor, characterized in that, Applied to the server side, the method includes: According to a preset cycle, the video sequence inside the reactor to be analyzed is collected by a 5G explosion-proof camera pre-deployed in the reactor during the current cycle; The video sequence inside the vessel to be analyzed is input into a pre-trained chemical product visual analysis model, which outputs the current product visual information corresponding to the video data inside the vessel to be analyzed. The process involves: acquiring real-time reaction condition parameters of the reactor within the current cycle; combining the real-time reaction condition parameters and the current product visual information to obtain multiple real-time process quantification parameters within the current cycle; combining the chemical product visual description parameters and chemical reaction conditions of each process operation node to obtain multiple raw process quantification parameters for each process operation node; performing feature normalization on the multiple real-time process quantification parameters to obtain a first normalized feature; creating a master node and associating the first normalized feature with the master node to obtain a master node feature; and performing feature normalization on the multiple raw process quantification parameters of each process operation node to obtain... The second normalized feature of each process operation node; creating child nodes for each process operation node; associating the second normalized feature of each process operation node with the child nodes of each process operation node to obtain multiple child node features; quantifying the spatial distance between the main node feature and each child node feature to obtain the comprehensive matching degree of each process operation node; selecting the process operation node with the highest comprehensive matching degree as the target process operation node to be executed in the current cycle; preset process operation node information to reflect the operation steps to be executed in different process operation nodes, chemical product visual description parameters, and chemical reaction conditions; If the target process operation node is inconsistent with the historical process operation node of the previous historical cycle, an operation process prompt message for the target process operation node is created and sent to the client.

2. The method according to claim 1, characterized in that, The multiple real-time process quantification parameters include the RGB value of the current material color, the current temperature value inside the reactor, the current area ratio of the precipitate, and the current bubble generation frequency; the multiple original process quantification parameters include preset color range, preset temperature range, preset precipitate area range, and preset bubble generation frequency range. Based on the multiple real-time process quantization parameters and the multiple original process quantization parameters of each process operation node, a comprehensive matching degree is generated for each process operation node, including: Calculate the color matching degree between the RGB value of the current material color and the preset color range of each process operation node; Determine whether the current temperature value inside the reactor is within the preset temperature range of each process operation node, and generate a temperature matching result; Calculate the degree of overlap between the current precipitate area ratio and the preset precipitate area range for each process operation node to obtain the precipitate area matching degree; Determine whether the current bubble generation frequency is within the preset bubble generation frequency range of each process operation node, and generate a bubble frequency matching result; Determine the color matching degree, the temperature matching result, the sedimentation area matching degree, and the bubble frequency matching result; The matching degree of the color matching degree, the temperature matching result, the sedimentation area matching degree, and the bubble frequency matching result are calculated to obtain the comprehensive matching degree of each process operation node.

3. The method according to claim 1, characterized in that, The method further includes: If the target process operation node is consistent with the historical process operation node of the previous historical cycle, the step of collecting the video sequence inside the reactor to be analyzed in the current cycle by using a 5G explosion-proof camera pre-deployed in the reactor according to a preset cycle continues.

4. The method according to claim 1, characterized in that, The process of generating a pre-trained visual analysis model for chemical products, following these steps, also includes: Acquire historical video sequences collected from the historical operation process of the reactor, the historical video sequences being used to reflect visual information of the entire historical reaction cycle of the reactor; Analyze the historical time period of each historical process operation node from the entire historical reaction cycle; From the historical video sequence, extract the operation process video segment corresponding to the historical time period of each historical process operation node, and use it as the operation process video segment of each historical process operation node; Based on the video clips of the operation process of each historical process operation node, model training samples are constructed; A visual analysis model for chemical products is created using neural networks; The training samples of the model are input into the visual analysis model of chemical products, the model loss value is output, and when the loss value reaches the minimum, the pre-trained visual analysis model of chemical products is generated.

5. The method according to claim 4, characterized in that, The analysis of the historical time period for each historical process operation node from the entire historical reaction cycle includes: The historical video sequence is extracted frame by frame in chronological order to obtain multiple original video frames, each carrying a timestamp. Based on the preset process operation node information, obtain the historical operation steps and historical chemical product visual description parameters for each historical process operation node; From the multiple original video frames, traverse and search for video frames that match the historical operation steps and historical chemical product visual description parameters of each historical process operation node, so as to determine the node start frame and node end frame of each historical process operation node. The historical time period between the timestamps carried by the node start frame and the node end frame of each historical process operation node is determined as the historical time period of each historical process operation node.

6. The method according to claim 5, characterized in that, The step of traversing and searching through the plurality of original video frames to find video frames that match the historical operation steps and historical chemical product visual description parameters of each historical process operation node includes: Based on the inter-frame difference algorithm, the pixel change rate between adjacent video frames in the plurality of original video frames is calculated; The original video frames with pixel change rates greater than a preset threshold are traversed and obtained to obtain the video frame sequence to be analyzed. For each video frame in the video frame sequence to be analyzed, define operation tool information and operation gesture information to obtain the first operation step information for each video frame to be analyzed. Analyze the visual description information of the first chemical product in each video frame of the video frame sequence to be analyzed; The first operation step information and the first chemical product visual description information of each video frame to be analyzed are correlated with the historical operation steps and historical chemical product visual description parameters of each historical process operation node to obtain the correlation analysis results between each video frame to be analyzed and each historical process operation node. Based on the correlation analysis results, video frames that are related to each historical process operation node are identified to be analyzed, and video frames that match the historical operation steps and historical chemical product visual description parameters of each historical process operation node are obtained.

7. The method according to claim 4, characterized in that, The step of constructing model training samples based on the video clips of the operation process of each historical process operation node includes: Each video frame is obtained from the operation process video clips of each historical process operation node; Calculate the historical RGB values ​​of the material color, the historical area ratio of the sediment, and the historical bubble generation frequency in each video frame; The historical RGB values, the historical area ratio of sediment, and the historical bubble generation frequency are used as tag data; The labeled data is used to annotate each video frame to obtain model training samples.

8. A process monitoring device for an internal reactor implemented using the method described in any one of claims 1-7, characterized in that, The device includes: The video acquisition module is used to acquire the video sequence inside the reactor to be analyzed within the current cycle using a 5G explosion-proof camera pre-deployed in the reactor according to a preset cycle; The product visual information output module is used to input the video sequence inside the vessel to be analyzed into a pre-trained chemical product visual analysis model and output the current product visual information corresponding to the video data inside the vessel to be analyzed. The process operation node analysis module is used to analyze the target process operation node to be executed in the current cycle based on the preset process operation node information and the current product visual information; the preset process operation node information is used to reflect the operation steps to be executed in different process operation nodes, the chemical product visual description parameters, and the chemical reaction conditions. The operation process prompt information creation module is used to create operation process prompt information for the target process operation node and send it to the client when the target process operation node is inconsistent with the historical process operation node of the previous historical cycle.

Images